The School of Computer Science at the University of Windsor is pleased to present…
Security of AI-enabled Multimodal Perception Systems
Colloquium Presentation by: Dr. Yi Zhu
Date: Friday, 23 October 2026
Time: 10:00 – 11:00 am
Location: Erie Hall, room 3123
Abstract:
AI-enabled perception systems are becoming a fundamental component of many intelligent and autonomous
systems, including autonomous vehicles, robotics, intelligent transportation, and other cyber-physical systems.
These perception systems increasingly rely on multiple sensing modalities such as cameras, LiDAR, radar, and
other sensors to understand complex physical environments and support safety-critical decision making.
However, some attackers may perform malicious attacks to compromise perception results, potentially causing
incorrect or unsafe system decisions. Such attacks raise significant concerns about the security, reliability, and
trustworthiness of AI-enabled perception systems, particularly when they are deployed in safety-critical
applications such as autonomous driving. In this talk, I will first examine security vulnerabilities and malicious
attacks targeting individual sensing modalities, with a particular focus on LiDAR and radar. I will then present our
recent research on attacking multimodal, multi-sensor fusion-based perception systems that integrate camera,
LiDAR, and radar data, highlighting how vulnerabilities can propagate across modalities and affect downstream
perception tasks. Finally, I will discuss open research challenges and future directions toward building secure,
robust, and trustworthy Physical AI.
Keywords: Multimodal Perception, AI Security, Adversarial Attacks and Defenses.
Biography:
Yi Zhu is an Assistant Professor of the Department of Computer Science at Wayne State University. Before that, he
obtained his Ph.D. in 2024 from the Department of Computer Science and Engineering, University at Buffalo. His
research interests lie in the broad areas of Physical AI and Security & Privacy. The primary goal of his research is to
build secure and trustworthy AI for high-stakes domains such as autonomous driving and smart manufacturing.
His research outcomes have been published in various top venues such as CCS, NDSS, USENIX Security,
MobiCom and SenSys.
